Diatom-based transfer functions for inferring past hydrochemical characteristics of African lakes

Abstract

Abstract A new dataset of 282 modern diatom samples and associated environmental information has been created by merging existing regional datasets from North and East Africa and Niger. The relationships between diatom species distributions and hydrochemistry are examined using canonical correspondence analysis (CCA) and partial CCA. Variables reflecting water conductivity, pH, and cation and anion composition account for significant and independent components of the total variation in the diatom data. Predictive models (transfer functions) are developed using the method of weighted averaging for conductivity ( r 2 = 0.87), pH ( r 2 = 0.77), and ratios between alkali and alkaline earth metals ( r 2 = 0.81), and carbonate-bicarbonate and sulphate + chloride ions ( r 2 = 0.82). Prediction errors are estimated using the computer-intensive method of jackknifing. These transfer functions enlarge the potential domain for reconstruction of past hydrochemistry from fossil diatoms preserved in lake sediments.